references/lobe-chat
2
0
Fork 0
mirror of https://github.com/lobehub/lobe-chat.git synced 2026-06-16 12:36:07 +00:00
Code Issues Packages Projects Releases Wiki Activity

✨ feat: add online search support for available providers (#6475)

Browse Source 
* ✨ feat: add online search support for available providers

* 🔨 chore: rollback changes for minimax
This commit is contained in:
Zhijie He
2025-02-27 11:51:10 +08:00
committed by GitHub
parent 257a87e664
commit cb0a3bc2fc
15 changed files with 302 additions and 30 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/config/aiModels/ai360.ts
+4
View File
@@ -52,6 +52,7 @@ const ai360ChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 8000,
description:
@@ -64,6 +65,9 @@ const ai360ChatModels: AIChatModelCard[] = [
input: 2,
output: 5,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
src/config/aiModels/baichuan.ts
+16
View File
@@ -4,6 +4,7 @@ const baichuanChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_768,
description:
@@ -17,11 +18,15 @@ const baichuanChatModels: AIChatModelCard[] = [
input: 100,
output: 100,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_768,
description:
@@ -35,11 +40,15 @@ const baichuanChatModels: AIChatModelCard[] = [
input: 15,
output: 15,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_768,
description:
@@ -53,11 +62,15 @@ const baichuanChatModels: AIChatModelCard[] = [
input: 0.98,
output: 0.98,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_768,
description:
@@ -70,6 +83,9 @@ const baichuanChatModels: AIChatModelCard[] = [
input: 12,
output: 12,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
src/config/aiModels/hunyuan.ts
+40
View File
@@ -20,6 +20,9 @@ const hunyuanChatModels: AIChatModelCard[] = [
type: 'chat',
},
{
abilities: {
search: true,
},
contextWindowTokens: 32_000,
description:
'采用更优的路由策略，同时缓解了负载均衡和专家趋同的问题。长文方面，大海捞针指标达到99.9%。MOE-32K 性价比相对更高，在平衡效果、价格的同时，可对实现对长文本输入的处理。',
@@ -33,9 +36,15 @@ const hunyuanChatModels: AIChatModelCard[] = [
output: 2,
},
releasedAt: '2025-02-10',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
search: true,
},
contextWindowTokens: 256_000,
description:
'采用更优的路由策略，同时缓解了负载均衡和专家趋同的问题。长文方面，大海捞针指标达到99.9%。MOE-256K 在长度和效果上进一步突破，极大的扩展了可输入长度。',
@@ -49,11 +58,15 @@ const hunyuanChatModels: AIChatModelCard[] = [
output: 2,
},
releasedAt: '2025-02-10',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_000,
description:
@@ -68,11 +81,15 @@ const hunyuanChatModels: AIChatModelCard[] = [
output: 50,
},
releasedAt: '2025-01-10',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_000,
description:
@@ -86,11 +103,15 @@ const hunyuanChatModels: AIChatModelCard[] = [
output: 50,
},
releasedAt: '2025-01-10',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_000,
description:
@@ -104,11 +125,15 @@ const hunyuanChatModels: AIChatModelCard[] = [
output: 50,
},
releasedAt: '2025-01-10',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_000,
description:
@@ -122,9 +147,15 @@ const hunyuanChatModels: AIChatModelCard[] = [
output: 50,
},
releasedAt: '2024-11-20',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
search: true,
},
contextWindowTokens: 32_000,
description:
'Hunyuan-large 模型总参数量约 389B，激活参数量约 52B，是当前业界参数规模最大、效果最好的 Transformer 架构的开源 MoE 模型。',
@@ -138,9 +169,15 @@ const hunyuanChatModels: AIChatModelCard[] = [
output: 12,
},
releasedAt: '2025-02-10',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
search: true,
},
contextWindowTokens: 134_000,
description:
'擅长处理长文任务如文档摘要和文档问答等，同时也具备处理通用文本生成任务的能力。在长文本的分析和生成上表现优异，能有效应对复杂和详尽的长文内容处理需求。',
@@ -154,6 +191,9 @@ const hunyuanChatModels: AIChatModelCard[] = [
output: 18,
},
releasedAt: '2024-12-18',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
src/config/aiModels/moonshot.ts
+32
View File
@@ -4,6 +4,7 @@ const moonshotChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
search: true,
vision: true,
},
contextWindowTokens: 131_072,
@@ -19,11 +20,15 @@ const moonshotChatModels: AIChatModelCard[] = [
output: 60,
},
releasedAt: '2025-02-17',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 131_072,
description:
@@ -35,11 +40,15 @@ const moonshotChatModels: AIChatModelCard[] = [
input: 60,
output: 60,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 8192,
description:
@@ -51,11 +60,15 @@ const moonshotChatModels: AIChatModelCard[] = [
input: 12,
output: 12,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_768,
description:
@@ -67,11 +80,15 @@ const moonshotChatModels: AIChatModelCard[] = [
input: 24,
output: 24,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 131_072,
description:
@@ -83,11 +100,15 @@ const moonshotChatModels: AIChatModelCard[] = [
input: 60,
output: 60,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
vision: true,
},
contextWindowTokens: 8192,
@@ -101,11 +122,15 @@ const moonshotChatModels: AIChatModelCard[] = [
output: 12,
},
releasedAt: '2025-01-14',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
vision: true,
},
contextWindowTokens: 32_768,
@@ -119,11 +144,15 @@ const moonshotChatModels: AIChatModelCard[] = [
output: 24,
},
releasedAt: '2025-01-14',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
vision: true,
},
contextWindowTokens: 131_072,
@@ -137,6 +166,9 @@ const moonshotChatModels: AIChatModelCard[] = [
output: 60,
},
releasedAt: '2025-01-14',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
];
src/config/aiModels/spark.ts
+30
View File
@@ -2,6 +2,9 @@ import { AIChatModelCard } from '@/types/aiModel';
const sparkChatModels: AIChatModelCard[] = [
{
abilities: {
search: true,
},
contextWindowTokens: 8192,
description:
'Spark Lite 是一款轻量级大语言模型，具备极低的延迟与高效的处理能力，完全免费开放，支持实时在线搜索功能。其快速响应的特性使其在低算力设备上的推理应用和模型微调中表现出色，为用户带来出色的成本效益和智能体验，尤其在知识问答、内容生成及搜索场景下表现不俗。',
@@ -9,9 +12,15 @@ const sparkChatModels: AIChatModelCard[] = [
enabled: true,
id: 'lite',
maxOutput: 4096,
settings: {
searchImpl: 'internal',
},
type: 'chat',
},
{
abilities: {
search: true,
},
contextWindowTokens: 8192,
description:
'Spark Pro 是一款为专业领域优化的高性能大语言模型，专注数学、编程、医疗、教育等多个领域，并支持联网搜索及内置天气、日期等插件。其优化后模型在复杂知识问答、语言理解及高层次文本创作中展现出色表现和高效性能，是适合专业应用场景的理想选择。',
@@ -19,9 +28,15 @@ const sparkChatModels: AIChatModelCard[] = [
enabled: true,
id: 'generalv3',
maxOutput: 8192,
settings: {
searchImpl: 'internal',
},
type: 'chat',
},
{
abilities: {
search: true,
},
contextWindowTokens: 131_072,
description:
'Spark Pro 128K 配置了特大上下文处理能力，能够处理多达128K的上下文信息，特别适合需通篇分析和长期逻辑关联处理的长文内容，可在复杂文本沟通中提供流畅一致的逻辑与多样的引用支持。',
@@ -29,11 +44,15 @@ const sparkChatModels: AIChatModelCard[] = [
enabled: true,
id: 'pro-128k',
maxOutput: 4096,
settings: {
searchImpl: 'internal',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 8192,
description:
@@ -42,11 +61,15 @@ const sparkChatModels: AIChatModelCard[] = [
enabled: true,
id: 'generalv3.5',
maxOutput: 8192,
settings: {
searchImpl: 'internal',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_768,
description:
@@ -55,11 +78,15 @@ const sparkChatModels: AIChatModelCard[] = [
enabled: true,
id: 'max-32k',
maxOutput: 8192,
settings: {
searchImpl: 'internal',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 8192,
description:
@@ -68,6 +95,9 @@ const sparkChatModels: AIChatModelCard[] = [
enabled: true,
id: '4.0Ultra',
maxOutput: 8192,
settings: {
searchImpl: 'internal',
},
type: 'chat',
},
];
src/config/aiModels/stepfun.ts
+44
View File
@@ -6,6 +6,7 @@ const stepfunChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 8000,
description: '高速模型，适合实时对话。',
@@ -17,11 +18,15 @@ const stepfunChatModels: AIChatModelCard[] = [
input: 1,
output: 4,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 8000,
description: '小型模型，适合轻量级任务。',
@@ -33,11 +38,15 @@ const stepfunChatModels: AIChatModelCard[] = [
input: 5,
output: 20,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 32_000,
description: '支持中等长度的对话，适用于多种应用场景。',
@@ -49,11 +58,15 @@ const stepfunChatModels: AIChatModelCard[] = [
input: 15,
output: 70,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 128_000,
description: '平衡性能与成本，适合一般场景。',
@@ -65,11 +78,15 @@ const stepfunChatModels: AIChatModelCard[] = [
input: 40,
output: 200,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 256_000,
description: '具备超长上下文处理能力，尤其适合长文档分析。',
@@ -80,11 +97,15 @@ const stepfunChatModels: AIChatModelCard[] = [
input: 95,
output: 300,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 16_000,
description: '支持大规模上下文交互，适合复杂对话场景。',
@@ -96,11 +117,15 @@ const stepfunChatModels: AIChatModelCard[] = [
input: 38,
output: 120,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 8000,
description:
@@ -114,11 +139,15 @@ const stepfunChatModels: AIChatModelCard[] = [
output: 2,
},
releasedAt: '2025-01-14',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 16_000,
description: 'step-2模型的实验版本，包含最新的特性，滚动更新中。不推荐在正式生产环境使用。',
@@ -131,11 +160,15 @@ const stepfunChatModels: AIChatModelCard[] = [
output: 120,
},
releasedAt: '2025-01-15',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
vision: true,
},
contextWindowTokens: 8000,
@@ -148,11 +181,15 @@ const stepfunChatModels: AIChatModelCard[] = [
input: 5,
output: 20,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
vision: true,
},
contextWindowTokens: 32_000,
@@ -164,11 +201,15 @@ const stepfunChatModels: AIChatModelCard[] = [
input: 15,
output: 70,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
vision: true,
},
contextWindowTokens: 32_000,
@@ -182,6 +223,9 @@ const stepfunChatModels: AIChatModelCard[] = [
output: 70,
},
releasedAt: '2025-01-22',
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
src/config/aiModels/zhipu.ts
+36
View File
@@ -20,6 +20,7 @@ const zhipuChatModels: AIChatModelCard[] = [
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 128_000,
description: 'GLM-4-Flash 是处理简单任务的理想选择，速度最快且免费。',
@@ -31,11 +32,15 @@ const zhipuChatModels: AIChatModelCard[] = [
input: 0,
output: 0,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 128_000,
description: 'GLM-4-FlashX 是Flash的增强版本，超快推理速度。',
@@ -47,11 +52,15 @@ const zhipuChatModels: AIChatModelCard[] = [
input: 0.1,
output: 0.1,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 1_024_000,
description: 'GLM-4-Long 支持超长文本输入，适合记忆型任务与大规模文档处理。',
@@ -62,11 +71,15 @@ const zhipuChatModels: AIChatModelCard[] = [
input: 1,
output: 1,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 128_000,
description: 'GLM-4-Air 是性价比高的版本，性能接近GLM-4，提供快速度和实惠的价格。',
@@ -78,11 +91,15 @@ const zhipuChatModels: AIChatModelCard[] = [
input: 1,
output: 1,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 8192,
description: 'GLM-4-AirX 提供 GLM-4-Air 的高效版本，推理速度可达其2.6倍。',
@@ -94,11 +111,15 @@ const zhipuChatModels: AIChatModelCard[] = [
input: 10,
output: 10,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 128_000,
description:
@@ -110,11 +131,15 @@ const zhipuChatModels: AIChatModelCard[] = [
input: 100,
output: 100,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 128_000,
description: 'GLM-4-Plus 作为高智能旗舰，具备强大的处理长文本和复杂任务的能力，性能全面提升。',
@@ -126,11 +151,15 @@ const zhipuChatModels: AIChatModelCard[] = [
input: 50,
output: 50,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 128_000,
description: 'GLM-4-0520 是最新模型版本，专为高度复杂和多样化任务设计，表现卓越。',
@@ -141,11 +170,15 @@ const zhipuChatModels: AIChatModelCard[] = [
input: 100,
output: 100,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
abilities: {
functionCall: true,
search: true,
},
contextWindowTokens: 128_000,
description: 'GLM-4 是发布于2024年1月的旧旗舰版本，目前已被更强的 GLM-4-0520 取代。',
@@ -156,6 +189,9 @@ const zhipuChatModels: AIChatModelCard[] = [
input: 100,
output: 100,
},
settings: {
searchImpl: 'params',
},
type: 'chat',
},
{
src/libs/agent-runtime/ai360/index.ts
+15 -2
View File
@@ -13,9 +13,22 @@ export const LobeAi360AI = LobeOpenAICompatibleFactory({
baseURL: 'https://api.360.cn/v1',
chatCompletion: {
handlePayload: (payload) => {
const { enabledSearch, tools, ...rest } = payload;
const ai360Tools = enabledSearch ? [
...(tools || []),
{
type: "web_search",
web_search: {
search_mode: "auto",
},
}
] : tools;
return {
...payload,
stream: !payload.tools,
...rest,
stream: !ai360Tools,
tools: ai360Tools,
} as any;
},
},
src/libs/agent-runtime/baichuan/index.ts
+14 -4
View File
@@ -1,5 +1,3 @@
import OpenAI from 'openai';
import { ChatStreamPayload, ModelProvider } from '../types';
import { LobeOpenAICompatibleFactory } from '../utils/openaiCompatibleFactory';
@@ -17,14 +15,26 @@ export const LobeBaichuanAI = LobeOpenAICompatibleFactory({
baseURL: 'https://api.baichuan-ai.com/v1',
chatCompletion: {
handlePayload: (payload: ChatStreamPayload) => {
const { temperature, ...rest } = payload;
const { enabledSearch, temperature, tools, ...rest } = payload;
const baichuanTools = enabledSearch ? [
...(tools || []),
{
type: "web_search",
web_search: {
enable: true,
search_mode: process.env.BAICHUAN_SEARCH_MODE || "performance_first", // performance_first or quality_first
},
}
] : tools;
return {
...rest,
// [baichuan] frequency_penalty must be between 1 and 2.
frequency_penalty: undefined,
temperature: temperature !== undefined ? temperature / 2 : undefined,
} as OpenAI.ChatCompletionCreateParamsStreaming;
tools: baichuanTools,
} as any;
},
},
debug: {
src/libs/agent-runtime/hunyuan/index.ts
+18
View File
@@ -9,6 +9,24 @@ export interface HunyuanModelCard {
export const LobeHunyuanAI = LobeOpenAICompatibleFactory({
baseURL: 'https://api.hunyuan.cloud.tencent.com/v1',
chatCompletion: {
handlePayload: (payload) => {
const { enabledSearch, ...rest } = payload;
return {
...rest,
...(enabledSearch && {
/*
citation: true,
enable_multimedia: true,
search_info: true
*/
enable_enhancement: true,
enable_speed_search: process.env.HUNYUAN_ENABLE_SPEED_SEARCH === '1',
}),
} as any;
},
},
debug: {
chatCompletion: () => process.env.DEBUG_HUNYUAN_CHAT_COMPLETION === '1',
},
src/libs/agent-runtime/minimax/index.ts
+2 -11
View File
@@ -12,23 +12,14 @@ export const LobeMinimaxAI = LobeOpenAICompatibleFactory({
baseURL: 'https://api.minimax.chat/v1',
chatCompletion: {
handlePayload: (payload) => {
const { temperature, top_p, ...params } = payload;
const { max_tokens, temperature, top_p, ...params } = payload;
return {
...params,
frequency_penalty: undefined,
max_tokens: payload.max_tokens !== undefined ? payload.max_tokens : getMinimaxMaxOutputs(payload.model),
max_tokens: max_tokens !== undefined ? max_tokens : getMinimaxMaxOutputs(payload.model),
presence_penalty: undefined,
stream: true,
temperature: temperature === undefined || temperature <= 0 ? undefined : temperature / 2,
tools: params.tools?.map((tool) => ({
function: {
description: tool.function.description,
name: tool.function.name,
parameters: JSON.stringify(tool.function.parameters),
},
type: 'function',
})),
top_p: top_p !== undefined && top_p > 0 && top_p <= 1 ? top_p : undefined,
} as any;
},
src/libs/agent-runtime/moonshot/index.ts
+13 -4
View File
@@ -1,5 +1,3 @@
import OpenAI from 'openai';
import { ChatStreamPayload, ModelProvider } from '../types';
import { LobeOpenAICompatibleFactory } from '../utils/openaiCompatibleFactory';
@@ -13,12 +11,23 @@ export const LobeMoonshotAI = LobeOpenAICompatibleFactory({
baseURL: 'https://api.moonshot.cn/v1',
chatCompletion: {
handlePayload: (payload: ChatStreamPayload) => {
const { temperature, ...rest } = payload;
const { enabledSearch, temperature, tools, ...rest } = payload;
const moonshotTools = enabledSearch ? [
...(tools || []),
{
function: {
name: "$web_search",
},
type: "builtin_function",
}
] : tools;
return {
...rest,
temperature: temperature !== undefined ? temperature / 2 : undefined,
} as OpenAI.ChatCompletionCreateParamsStreaming;
tools: moonshotTools,
} as any;
},
},
debug: {
src/libs/agent-runtime/qwen/index.ts
+4
View File
@@ -50,6 +50,10 @@ export const LobeQwenAI = LobeOpenAICompatibleFactory({
...(enabledSearch && {
enable_search: enabledSearch,
search_options: {
/*
enable_citation: true,
enable_source: true,
*/
search_strategy: process.env.QWEN_SEARCH_STRATEGY || 'standard', // standard or pro
},
}),
src/libs/agent-runtime/stepfun/index.ts
+15 -2
View File
@@ -11,9 +11,22 @@ export const LobeStepfunAI = LobeOpenAICompatibleFactory({
baseURL: 'https://api.stepfun.com/v1',
chatCompletion: {
handlePayload: (payload) => {
const { enabledSearch, tools, ...rest } = payload;
const stepfunTools = enabledSearch ? [
...(tools || []),
{
function: {
description: "use web_search to search information on the internet",
},
type: "web_search",
}
] : tools;
return {
...payload,
stream: !payload.tools,
...rest,
stream: !stepfunTools,
tools: stepfunTools,
} as any;
},
},
src/libs/agent-runtime/zhipu/index.ts
+19 -7
View File
@@ -1,6 +1,4 @@
import OpenAI from 'openai';
import { ChatStreamPayload, ModelProvider } from '../types';
import { ModelProvider } from '../types';
import { LobeOpenAICompatibleFactory } from '../utils/openaiCompatibleFactory';
import type { ChatModelCard } from '@/types/llm';
@@ -14,9 +12,21 @@ export interface ZhipuModelCard {
export const LobeZhipuAI = LobeOpenAICompatibleFactory({
baseURL: 'https://open.bigmodel.cn/api/paas/v4',
chatCompletion: {
handlePayload: ({ max_tokens, model, temperature, top_p, ...payload }: ChatStreamPayload) =>
({
...payload,
handlePayload: (payload) => {
const { enabledSearch, max_tokens, model, temperature, tools, top_p, ...rest } = payload;
const zhipuTools = enabledSearch ? [
...(tools || []),
{
type: "web_search",
web_search: {
enable: true,
},
}
] : tools;
return {
...rest,
max_tokens:
max_tokens === undefined ? undefined :
(model.includes('glm-4v') && Math.min(max_tokens, 1024)) ||
@@ -24,6 +34,7 @@ export const LobeZhipuAI = LobeOpenAICompatibleFactory({
max_tokens,
model,
stream: true,
tools: zhipuTools,
...(model === 'glm-4-alltools'
? {
temperature:
@@ -36,7 +47,8 @@ export const LobeZhipuAI = LobeOpenAICompatibleFactory({
temperature: temperature !== undefined ? temperature / 2 : undefined,
top_p,
}),
}) as OpenAI.ChatCompletionCreateParamsStreaming,
} as any;
},
},
constructorOptions: {
defaultHeaders: {